Re: [ADMIN] Speed problem - Mailing list pgsql-admin

From Gene Selkov Jr.
Subject Re: [ADMIN] Speed problem
Date
Msg-id 199811051850.MAA28215@antares.mcs.anl.gov
Whole thread Raw
In response to Re: [ADMIN] Speed problem  (Roger Books <books@mail.state.fl.us>)
List pgsql-admin
> On the memory issue, I'm running PII-300's with 128MB of RAM.
> while memory is always full (this machine is also running bind)
> and has stuff swapped, there is actually very little swapping
> there is actually very little swapping going on.

Full memory does automatically mean swapping is going on. It may be
filled up with the filesystem cache, which is normally allowed to
grow as much as it can until other processes request for more memory.

> Maybe it's just me but I don't recall ever seeing
> a Unix box that didn't have something swapped.

Two other things to consider. One is that it is almost impossible
to see a Unix box not running Netscape (unless it is an unattended
sever). All linux versions of Netscape after 2.* had various memory
leaks. 4.*s do something wrong to the swapping, such that a lot of
swap space is not getting reclaimed even after you kill them. Netscape
in linux is also known to leave non-functional child processes in
memory. In short, if you are in linux, make sure Netscape hasn't
been run since the system is up, in order to examine memory problems.

Other thing is, you can't always trust information reported by ps.
The last time I had problems with a large COPY (also reported by a
number of others), my machine was swapping badly, memory was taken up
but ps, even though it listed postgres as the top process, reported
only 3Mb memory usage.

--Gene

pgsql-admin by date:

Previous
From: "John E. Garrott"
Date:
Subject: 6.4 installation
Next
From: Carl Riches
Date:
Subject: unable to compile postgresql-6.3.2 on alpha-dec-osf4.0